L-M神经网络的磨削淬硬参数预测

The prediction of parameters of grind-hardening based on the L-M neural network

针对基于传统BP算法的神经网络训练中收敛速度较慢的缺点,提出一种基于L-M(Levenberg-Marquardt)算法的磨削淬硬层厚度预测,并开发了基于L-M算法的磨削淬硬神经网络预测系统。仿真结果表明:该系统模型显著缩短了训练时间,具有较高的准确性。通过网络训练和网络检验,得出该神经网络系统的预测值与实测值十分接近的结论,可充分证明L-M法BP神经网络对于磨削淬硬参数预测具有很好的效果。

An improved neural network based on L-M algorithm has been applied to the prediction of the thickness of the grind-hardening layer against to the slow convergence rate of conventional BP neural network. And neural network prediction system for grind-hardening process has been developed based on L-M algorithm. Simulation results indicate that this model can remarkably reduce the training time ,with its relatively high accuracy. By training and testing the network ,it has been concluded that predicted and measured values are very close by using this network. So neural network method based on L-M algorithm has very good effect in the prediction of grinding hardening parameters.